Dual-task research has revealed a response-selection bottleneck: response selection cannot occur simultaneously in two different tasks, though perceptual processing may overlap response selection (Welford, 1952; Pashler & Johnston, 1989). In serial performance with preview, the same task is performed repeatedly, but stimuli are available well before response. Does the dual-task bottleneck limit the rate of responding in this situation, despite the fact that task set need not change? Four experiments examined manual responses to letters, varying stimulus preview. Preview increased RTs for the first response and reduce the subsequent inter-response intervals (IRIs). Preview also reduced the effect of visual intensity and response duration on IRIs, whereas effects of stimulus-response mapping variables were unchanged. These results indicate that the response-selection bottleneck limits serial performance, just as it limits concurrent performance of two unrelated tasks. This implies that the response-selection bottleneck is not caused by the need to switch task set.